New Era of Agentic AI
Jensen Huang, the CEO of Nvidia, recently sat down with Marc Benioff, the CEO of Salesforce, at the Dreamforce conference to discuss the transformative role of artificial intelligence (AI) in modern society. Huang passionately outlined how we are entering a new era where AI agents will be seamlessly integrated into our day-to-day tasks, fundamentally shifting our industry from tool-centric to skill-driven.
AI Agents: The New Workforce
For the first time, AI agents are poised to sit atop traditional tools, utilizing and collaborating with them to solve complex problems. These agents will not only comprehend the nuances of our requests but will also possess the ability to reason, collaborate with other AI agents, and even develop and test new tools. This marks the beginning of an “agentic” future, where AI capabilities will grow exponentially as core models and frameworks evolve.
The Breakthrough of Unsupervised Learning
Huang highlighted the revolutionary impact of unsupervised learning, a method that allows AI to learn without the need for human-labeled data. This advancement eradicates the bottleneck created by human intervention, enabling AI to rapidly scale its capabilities. He cited AlphaGo as a prime example, where AI achieved superhuman performance through self-play and iteration, all without human input.
Scaling Data: The Next Frontier
As all publicly available data has been utilized, AI now faces the challenge of doing more with existing data or generating synthetic data. Projects like OpenAI’s rumored Q* (later renamed 01) exemplify how AI can create data for other AI models. By removing humans as a limiting factor, the pace of AI advancement is set to accelerate, potentially leading to an intelligence explosion.
Surpassing Moore’s Law: The Role of Computing Power
The rapid advancement in computing power, surpassing Moore’s Law, has been instrumental in AI’s growth. The shift from CPUs to GPUs and the transition from human-engineered software to machine learning software have created a feedback loop where AI aids in the development of new AI. This dynamic interaction is pushing technological progress at an unprecedented rate.
Ensuring Safety and Accuracy in AI
Huang emphasized the critical importance of safety and guardrails in AI development. Techniques such as fine-tuning, supervised learning, and reflection through chains of thought are essential to ensure AI generates safe and accurate responses. As AI becomes more embedded in our lives, maintaining these safety measures is crucial.
Practical Applications and Business Integration
The conversation also touched on the practical application of AI in businesses. Benioff stressed the need for companies to understand and implement AI effectively, demystifying the technology to make it accessible for practical use. Building AI agents should be straightforward, not a complex, multi-year project.
Huang concurred, comparing the onboarding of AI agents in businesses to onboarding new employees. Providing the right context, training, and documentation will enable AI agents to perform tasks efficiently from the outset. As AI continues to advance, the process of integrating and collaborating with AI agents will become more streamlined.
A Dynamic Future of Computing
In summary, Huang expressed his excitement about the future of computing. He envisions a world where AI dynamically creates and predicts software, delivering what users need in real-time. This paradigm shift, driven by advancements in hardware and software, promises to revolutionize our interaction with technology.
The discussion at Dreamforce highlighted the immense potential and rapid evolution of AI. As we stand on the brink of this agentic future, the integration of AI into everyday tasks is set to transform industries and redefine the boundaries of what is possible.